LspCAD active xo queries

[tcpip]

Asking for help from all LspCAD veterans, specially those who have used its active filter facilities.

A few friends and I who do diy audio together were seriously thinking of investing in something which could do active filter optimisations (spl, phase), so that we could then build active xo as finely tuned as the passive xo done with CALSOD/ Speaker Workshop. We looked at LspCAD, which is the least expensive option and well regarded overall. But we found that some of its active filter building blocks have inductors. If you read the Ugly Duckling case study on the LspCAD Website, written by the author of LspCAD, you'll see one inductor in the schematic of his optimised active filter on page 30. One of my friends from our diy gang joined the LspCAD users mailing list and asked why this was so, but there's been no answer to the "why". Ingemar Johansson, the author of LspCAD, has replied very courteously saying that some of his active filter building blocks do indeed have inductors. And "we can replace each inductor by a gyrator ourselves".

Frankly, this was not a satisfactory state of affairs. We want a good active xo optimisation program without us having to replace inductors with gyrators. What's worse, those gyrators would be "outside" the software, which means any subtle errors introduced while converting each inductor to a gyrator would not show up in the modelling and simulation.

What is confusing me even more is: why were inductors necessary in the first place?

My questions to you:

• Are you comfortable building active filters with the odd inductor? Do you think we're being too paranoid?
• Do all software packages handling active xo design and optimisation force one to use inductors this way? What about SoundEasy? (That was the next option on our software list.)
• What software package would you recommend for active xo optimisation, together with all the other features that Speaker Workshop and LspCAD has? (LspCAD scores over Speaker Workshop even in passive xo, because it does global optimisation of all branches of the xo.)

Thanks for any help and suggestions.

 

[ergo]

tcpip,

I'm afraid you are missing a point here a little.

Active filters in LspCad are just as they are usually - only R and C + opamp.

The ones including inductors are Q boost and Q cut circuits - these are equal to your RLC series compensation networks in passive filters. You will only be needing them if you have a driver with a hard cone breakup resonance etc.

As you know you can not do a resonant circuit just with R and C. If you need such boost or cut circuit you'll have to accept either using inductor or gyrator. There is no other way.

Actually the other way would be to design a filter so that you need only standard filter blocks

So I think your claims for Ingemar not helping you are a bit rough, He answered very politely and gave you an absolutely correct answer.

See attached pic. This is the sample of boost circuit LspCad uses.

Ergo

 

[tcpip]

Active filters in LspCad are just as they are usually - only R and C + opamp.

The ones including inductors are Q boost and Q cut circuits - these are equal to your RLC series compensation networks in passive filters. You will only be needing them if you have a driver with a hard cone breakup resonance etc.

As you know you can not do a resonant circuit just with R and C. If you need such boost or cut circuit you'll have to accept either using inductor or gyrator. There is no other way.

Thanks for your answer. This has been very helpful. Actually, I didn't know you cannot do a resonant circuit using just R and C, and I didn't know that LspCAD's usage of L is only in those Q-boost and Q-cut circuits. (Our gang comprisers beginners.) If someone had told me or my friends this earlier on the LspCAD mailing list, we'd have understood the precise L exposure much more clearly. That's why your response has been a big help.

Actually the other way would be to design a filter so that you need only standard filter blocks

Given my limited knowledge of active filters, I'd have tried implementing the Q-boost and Q-cut filters using conventional bandpass filters (for Q-boost) and notch filters (for cut). Would such an approach work, do you think? They would of course need more components than an LC-based resonant circuit.

So I think your claims for Ingemar not helping you are a bit rough, He answered very politely and gave you an absolutely correct answer.

I am sorry if that is what it seems like, because that was not my intention. But I feel I was 100% accurate. I have in fact mentioned that Ingemar did indeed reply to my friend and was very courteous, but he didn't say why inductors were needed in active filters.

From my earlier post:
Ingemar Johansson, the author of LspCAD, has replied very courteously saying that some of his active filter building blocks do indeed have inductors. And "we can replace each inductor by a gyrator ourselves".

If you want, I can quote the email that Ingemar sent my friend on the list. The key thing missing from Ingemar's post was what you've so helpfully explained in your post here, about there being no easy way to do resonant circuits using RC and opamp.

Thanks a lot for the help.

One more question, since you've been so patient till now: do all other active xo packages also ask you to use L for their Q-cut/boost filters? I guess they would, if I've understood your explanations right.

 

[ergo]

Using bandbass circuit works also if you do not have very sharp dip or peak.

Look at my project at
http://www.hifiklubi.ee/fat_margaret_project/measure.html

There you can see the woofer response for Seas W17E002. This driver has so strong cone resonants at 4.5 kHz that it has to be addressed in crossover as well. But I do not think it could be achieved with bandbass filters because of the very sharp and high Q notch is needed. Or if then it would add extreme complexity and still not sound good in the end.

****

As for other software packages, I haven't used too many besides LspCad. For hobby use the better software still quite expensive and I chose LspCad as the one to buy (great support available was one key plusses)

I just short looked at SoundEasy. I have a friends copy just to play around a little. But as I have discovered it's quite impossible to just play around SoundEasy. Even though there is huge number of options and capabilities, without reading the manual and studing everything slowly it's quite impossible to get any idea of the program.

Anyway SoundEasy seemed the have opamps and the regular passive R L C elements and one has to draw the whole schematic of active filter adding components one by one.

I think LspCad is hugely more easy to use even if it lacks some features others have. I'm at least very happy with LspCad

Ergo

 

[tcpip]

Thanks for your reply, Ergo. I really learned quite a bit from your Fat Margaret project pages. Very interesting enclosure design, and quite an interesting xo design description too.

And your kind of inputs was what I was looking for... someone who has really used LspCAD in one of his own projects. Thanks a lot. And yes, SoundEasy seems much more complex. If they had given a demo version, I'd have tried that to get a clearer picture.

 

[Davey]

tcpip,

Only notch/peak filters require inductors or active inductors and implementing them in a real circuit is not exactly trivial. There is some "optimization" usually required to fine tune the frequency, and dip/boost to get what you want because of component tolerances, parasitic effects, etc.

This portion of LspCAD is for fairly advanced designers and I believe Ingemar is making that assumption. Some further development is probably on his to-do list, but it's probably near the bottom and that's understandable.


See Siegfried's web page for some further information on how to implement active inductors:

http://www.woodartistry.com/linkwitzlab/filters.htm#7

Anyway, nowadays many people are switching to DSP-based crossovers. They take all the fun (and knowledge requirement) out of designing crosssovers like these.

Cheers,

Davey.